Data processing device that calculates an arrival probability for a destination using a user's movement history including a missing portion

• US 9,589,082 B2
• Filed: 11/07/2011
• Issued: 03/07/2017
• Est. Priority Date: 11/18/2010
• Status: Active Grant

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: Ide, Naoki, Ito, Masato, Sabe, Kohtaro
• Primary Examiner(s): Chad, Aniss
• Assistant Examiner(s): Rogers, David M
• Application Number:
  US13/878,920
• Publication Number:
  US 20130197890A1
• Time in Patent Office: 1,947 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G01C 21/3484 Personalized, e.g. from learned user behaviour or user-defined profiles
• CPC Class Codes: G01C 21/3617 using user history, behaviour, conditions or preferences, e.g. predicted or inferred from previous use or current movement
• CPC Class Codes: G01S 5/0009 Transmission of position information to remote stations involving assistance data G01S5/0236
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04
• CPC Class Codes: G06F 30/00 Computer-aided design [CAD]
• CPC Class Codes: G06F 30/20 Design optimisation, verification or simulation optimisation, verification or simulation of circuit designs G06F30/30
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G08G 1/0129 for creating historical data or processing based on historical data

Abstract

The present invention relates to a data processing device, a data processing method, and a program which enable prediction to be performed even when there is a gap in the current location data to be obtained in real time. A learning main processor 23 represents movement history data serving as data for learning, as a probability model which represents a user'"'"'s activity, and obtains a parameter thereof. A prediction main processor 33 uses the probability model obtained by learning to estimate a user'"'"'s current location from movement history data to be obtained in real time. In the event that there is a data missing portion included in movement history data to be obtained in real time, the prediction main processor 33 generates the data missing portion thereof by interpolation processing, and estimates state nose series corresponding to the interpolated data for prediction. With estimation of state node series, an observation probability less contribution of data than actual data is employed regarding interpolated data. The present invention may be applied to a data processing device configured to predict a destination from movement history data, for example.

Claims

1. A data processing device comprising:

circuitry configured to

obtain a parameter of a probability model when a user'"'"'s movement history data to be obtained as data for learning is represented as the probability model that represents the user'"'"'s activity;

estimate, of state nodes of the probability model using the parameter obtained by the circuitry, a destination node and a route point node equivalent to a movement destination and a route point;

obtain the user'"'"'s movement history data within a predetermined period of time from the present which differs from the data for learning, as data for prediction, and in the event that there is a data missing portion included in the obtained data for prediction, generate the data missing portion thereof by interpolation processing, and calculate virtual error with actual data corresponding to interpolated data generated by the interpolation processing;

input the data for prediction of which the data missing portion has been interpolated to the probability model using the parameter obtained by learning, and with estimation of state node series corresponding to the data for prediction of which the data missing portion has been interpolated, estimate a current point node equivalent to the user'"'"'s current location by using the virtual error as an observation probability of the state nodes regarding the interpolated data, and using an observation probability with less contribution of data than a full set of actual data;

determine whether the current point node obtained by the circuitry is a known-state node or an unknown-state node;

search, in the event that the current point node has been determined to be a known-state node, a route from the user'"'"'s current location to a destination using information regarding the destination node and the route point node and the current point node, and the probability model obtained by learning;

perform known/unknown determination based on an observation likelihood expected value of each of the state nodes to be calculated from the observation probability of the state nodes, and use the observation likelihood expected value to which a correction term that corrects the data missing portion so as to obtain a greater value than actual data is added;

calculate an arrival probability and time required for the searched destination;

and control a display or a speaker to output the arrival probability and the time required for the searched destination.

2. A data processing device comprising:

circuitry configured to

obtain a parameter of a probability model when a user'"'"'s movement history data to be obtained as data for learning is represented as the probability model that represents the user'"'"'s activity;

estimate, of state nodes of the probability model using the parameter obtained by the circuitry, a destination node and a route point node equivalent to a movement destination and a route point;

obtain the user'"'"'s movement history data within a predetermined period of time from the present which differs from the data for learning, as data for prediction, and in the event that there is a data missing portion included in the obtained data for prediction, generate the data missing portion thereof by interpolation processing, and calculate virtual error with actual data corresponding to interpolated data generated by the interpolation processing, the interpolation processing being linear interpolation on a past data missing portion not including a current point node, and interpolation processing by hold interpolation on a data missing portion including the current point node;

input the data for prediction of which the data missing portion has been interpolated to the probability model using the parameter obtained by learning, and with estimation of state node series corresponding to the data for prediction of which the data missing portion has been interpolated, estimate a current point node equivalent to the user'"'"'s current location by using the virtual error as an observation probability of the state nodes regarding the interpolated data, and using an observation probability with less contribution of data than a full set of actual data;

determine whether the current point node obtained by the circuitry is a known-state node or an unknown-state node;

search, in the event that the current point node has been determined to be a known-state node, a route from the user'"'"'s current location to a destination using information regarding the destination node and the route point node and the current point node, and the probability model obtained by learning;

perform known/unknown determination based on an observation likelihood expected value of each of the state nodes to be calculated from the observation probability of the state nodes, and use the observation likelihood expected value to which a correction term that corrects the data missing portion so as to obtain a greater value than actual data is added;

calculate an arrival probability and time required for the searched destination;

and control a display or a speaker to output the arrival probability and the time required for the searched destination.

3. The data processing device according to claim 2, wherein the circuitry is configured to calculate, regarding movement history data subjected to interpolation processing by the linear interpolation, error between interpolated data and smaller one of distance with movement history data immediately before data missing, or distance with movement history data immediately after data restoration, as the virtual error, and calculate, regarding movement data subjected to interpolation processing by the hold interpolation, the virtual error using a predetermined function according to elapsed time after data missing.

4. A data processing device comprising:

circuitry configured to

obtain a parameter of a probability model when a user'"'"'s movement history data to be obtained as data for learning is represented as the probability model that represents the user'"'"'s activity

by, in the event that after obtaining the parameter of the probability model once, movement history data serving as new data for learning is supplied, and there is a data missing portion included in the new data for learning, generating the data missing portion thereof by interpolation processing, and generating the new data for learning including the interpolated data, and determining whether the new data for learning including the interpolated data is movement history data in a known route or movement history data in an unknown route, and not performing updating of a probability distribution parameter of a state node of parameters of existing models which are the already obtained probability models regarding the interpolated data determined to be the movement history data in the known route;

estimate, of state nodes of the probability model using the parameter obtained by the circuitry, a destination node and a route point node equivalent to a movement destination and a route point;

obtain the user'"'"'s movement history data within a predetermined period of time from the present which differs from the data for learning as data for prediction and in the event that there is a data missing portion included in the obtained data for prediction, generate the data missing portion thereof by interpolation processing, and calculate virtual error with actual data corresponding to interpolated data generated by the interpolation processing;

input the data for prediction of which the data missing portion has been interpolated to the probability model using the parameter obtained by learning and with estimation of state node series corresponding to the data for prediction of which the data missing portion has been interpolated, estimate a current point node equivalent to the user'"'"'s current location by using the virtual error as an observation probability of the state nodes regarding the interpolated data, and using an observation probability with less contribution of data than a full set of actual data;

determine whether the current point node obtained by the circuitry is a known-state node or an unknown-state node;

search, in the event that the current point node has been determined to be a known-state node, a route from the user'"'"'s current location to a destination using information regarding the destination node and the route point node and the current point node, and the probability model obtained by learning;

perform known/unknown determination based on an observation likelihood expected value of each of the state nodes to be calculated from the observation probability of the state nodes, and use the observation likelihood expected value to which a correction term that corrects the data missing portion so as to obtain a greater value than actual data is added;

calculate an arrival probability and time required for the searched destination;

and

control a display or a speaker to output the arrival probability and the time required for the searched destination.

5. A data processing method comprising:

obtaining a parameter of a probability model when a user'"'"'s movement history data to be obtained as data for learning is represented as the probability model that represents the user'"'"'s activity;

estimating, of state nodes of the probability model using the parameter, a destination node and a route point node equivalent to a movement destination and a route point;

obtaining the user'"'"'s movement history data within a predetermined period of time from the present which differs from the data for learning, as data for prediction, and in the event that there is a data missing portion included in the obtained data for prediction, generating the data missing portion thereof by interpolation processing, and calculating virtual error with actual data corresponding to interpolated data generated by the interpolation processing;

inputting the data for prediction of which the data missing portion has been interpolated to the probability model using the parameter obtained by learning, and with estimation of state node series corresponding to the data for prediction of which the data missing portion has been interpolated, estimating a current point node equivalent to the user'"'"'s current location by using the virtual error as an observation probability of the state nodes regarding the interpolated data, and using an observation probability with less contribution of data than a full set of actual data;

determining whether the current point node is a known-state node or an unknown-state node;

searching, in the event that the current point node has been determined to be a known-state node, a route from the user'"'"'s current location to a destination using information regarding the destination node and the route point node and the current point node, and the probability model obtained by learning;

performing known/unknown determination based on an observation likelihood expected value of each of the state nodes to be calculated from the observation probability of the state nodes, and use the observation likelihood expected value to which a correction term that corrects the data missing portion so as to obtain a greater value than actual data is added;

calculating, using circuitry, an arrival probability and time required for the searched destination;

and controlling a display or a speaker to output the arrival probability and the time required for the searched destination.

6. A non-transitory computer readable medium including executable instructions which when executed by a computer cause the computer to:

obtain a parameter of a probability model when a user'"'"'s movement history data to be obtained as data for learning is represented as the probability model that represents the user'"'"'s activity;

estimate, of state nodes of the probability model using the parameter, a destination node and a route point node equivalent to a movement destination and a route point;

obtain the user'"'"'s movement history data within a predetermined period of time from the present which differs from the data for learning, as data for prediction, and in the event that there is a data missing portion included in the obtained data for prediction, generate the data missing portion thereof by interpolation processing, and calculate virtual error with actual data corresponding to interpolated data generated by the interpolation processing;

input the data for prediction of which the data missing portion has been interpolated to the probability model using the parameter obtained by learning, and with estimation of state node series corresponding to the data for prediction of which the data missing portion has been interpolated, estimate a current point node equivalent to the user'"'"'s current location by using the virtual error as an observation

probability of the state nodes regarding the interpolated data, and using an observation probability with less contribution of data than a full set of actual data;

determine whether the current point node is a known-state node or an unknown-state node;

search, in the event that the current point node has been determined to be a known-state node, a route from the user'"'"'s current location to a destination using information regarding the destination node and the route point node and the current point node, and the probability model obtained by learning;

perform known/unknown determination based on an observation likelihood expected value of each of the state nodes to be calculated from the observation probability of the state nodes, and use the observation likelihood expected value to which a correction term that corrects the data missing portion so as to obtain a greater value than actual data is added;

calculate an arrival probability and time required for the searched destination;

and control a display or a speaker to output the arrival probability and the time required for the searched destination.